The relationship between body mass index and per cent body fat in the severely obese

Background:  International standards define clinical obesity according to body mass index (BMI) without reference to age and gender. Recent studies among adults in the normal to mildly obese BMI ranges have shown that the relationship between BMI and per cent body fat (% fat) differs by age and gender. The extent to which age and gender affect the relationship between BMI and % fat among more severely obese individuals is less known. Aim:  The aim was to examine the age–gender association between measured BMI and % fat from a large cohort of adults, including a large number of severely obese subjects (1862 with a BMI ≥ 35 kg/m2). Methods:  BMI was computed from measured height and weight, and % fat was estimated from bioelectrical impedance in 3068 adults. Two impedance equations, the Sun equation and the Heath equation (specific to severe obesity), were used to calculate % fat. Results:  Average age for 991 men and 2077 women was 46 ± 15 vs. 44 ± 14 years respectively (p = 0.0003). The average BMI was 36 ± 9 kg/m2 for men and 39 ± 10 kg/m2 for women (p < 0.0001), with a combined gender BMI range of 19–74 kg/m2. Using the Sun equation, average % fat was 31 ± 8 vs. 46 ± 8% (p < 0.0001) for all men and women respectively. With the Sun equation, age-adjusted Spearman correlations between all BMI and % fat values were r = 0.80 and r = 0.83 for men and women, respectively, but only 0.60 (n = 479) and 0.61 (n = 1383) in severely obese participants (BMI ≥ 35 kg/m2). Using the Heath equation, only for participants with BMI ≥ 35 kg/m2, the age-adjusted Spearman correlations improved to r = 0.82 (n = 479) and r = 0.70 (n = 1383) for men and women respectively. Finally, by combining the Sun equation for subjects with BMI < 35 kg/m2 and the Heath equation for those with BMI ≥ 35 kg/m2, correlations improved to 0.89 for men and 0.87 for women. Using these combined equations, the relationship between BMI and % fat was best fit as a linear function for men and curvilinear function (both p < 0.001) for women across the range of BMI. The % fat was approximately 10% higher for any BMI value among women vs. men even among the severely obese (p < 0.0001). Conclusions:  These data that include a large cohort of severely obese individuals demonstrated a linear association between BMI and % fat for men and a curvilinear association between BMI and % fat for women when Sun and Heath equations were combined. Assuming disease risk is driven by adiposity, this study suggests a need to further explore the appropriateness of gender-specific BMI cutpoints for clinical risk assessment due to the marked difference in the BMI–per cent fat relation observed in men and women across the entire range of BMI.